refactor: remove AgenticLoopEngine and AGENT_LOOP from OPM

AgenticLoopEngine is now implemented in the standalone ovos-agentic-loop
repo (Agent Plugins/ovos-agentic-loop) and registers under the existing
opm.agents.chat entry-point group — no new OPM type is needed.

Removed:
- PluginTypes.AGENT_LOOP ("opm.agents.loop")
- PluginConfigTypes.AGENT_LOOP ("opm.agents.loop.config")
- AgenticLoopEngine class from templates/agents.py
- AgenticLoopEngine section from docs/api/agents.md

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: JarbasAl

docs/api/agents.md

@@ -63,34 +63,6 @@ Generate a response given a list of `AgentMessage` objects.
 
 ---
 
-### `AgenticLoopEngine`
-
-**Entry point:** `opm.agents.loop`
-
-A `ChatEngine` subclass for plugins that implement an internal agent loop (ReAct, tool-call/observe, background workers, etc.). Callers treat it identically to `ChatEngine` — the loop is an implementation detail.
-
-Adds a standard toolbox registration interface:
-
-```python
-def load_toolboxes(self, toolboxes: List[ToolBox]) -> None
-```
-
-Called by the persona loader when the persona config declares a `toolboxes` list. Plugins may also discover and load toolboxes internally.
-
-Persona config example:
-
-```json
-{
-  "name": "MyAgent",
-  "solvers": ["ovos-react-agent"],
-  "toolboxes": ["web_search_tools", "file_tools"]
-}
-```
-
-See [Agent Tools](agent-tools.md) for the `ToolBox` plugin API.
-
----
-
 ### `RetrievalEngine`
 
 **Entry point:** `opm.agents.retrieval`

ovos_plugin_manager/templates/agents.py

@@ -382,73 +382,6 @@ def get_response(self, utterance: str,
                                   units=units).content
 
 
-class AgenticLoopEngine(ChatEngine):
-    """
-    A ``ChatEngine`` subclass for plugins that implement an internal agent loop
-    (e.g. ReAct, tool-call/observe cycles, background worker agents).
-
-    From the perspective of a ``PersonaService`` or any caller, an
-    ``AgenticLoopEngine`` is identical to a ``ChatEngine`` — it receives a list
-    of ``AgentMessage`` objects and returns one.  All loop mechanics, tool
-    dispatch, retries, and background tasks are implementation details hidden
-    inside the plugin.
-
-    The ``toolboxes`` attribute gives OPM and persona configs a standard place
-    to inject ``ToolBox`` instances.  Plugins are free to discover and load
-    additional toolboxes internally.
-
-    Entry point group: ``opm.agents.loop``
-    """
-
-    def __init__(self, config: Optional[Dict[str, Any]] = None):
-        """
-        Initialise the engine and set up an empty toolbox registry.
-
-        Args:
-            config: Plugin-specific configuration dictionary.
-        """
-        super().__init__(config=config)
-        self.toolboxes: List[Any] = []  # List[ToolBox] — avoid circular import
-
-    def load_toolboxes(self, toolboxes: List[Any]) -> None:
-        """
-        Register a list of ``ToolBox`` instances with this engine.
-
-        Called by the persona loader when the persona config declares a
-        ``toolboxes`` list.  Plugins may also call this directly or discover
-        toolboxes on their own inside ``continue_chat``.
-
-        Args:
-            toolboxes: Instantiated ``ToolBox`` objects to make available to
-                the agent loop.
-        """
-        self.toolboxes = list(toolboxes)
-
-    @abc.abstractmethod
-    def continue_chat(self, messages: List[AgentMessage],
-                      session_id: str = "default",
-                      lang: Optional[str] = None,
-                      units: Optional[str] = None) -> AgentMessage:
-        """
-        Run the agent loop and return the final response.
-
-        The implementation is responsible for all internal steps: tool
-        selection, execution, observation, and iteration.  The caller always
-        receives a single ``AgentMessage`` with ``MessageRole.ASSISTANT``.
-
-        Args:
-            messages: Full conversation history up to and including the latest
-                user turn.
-            session_id: Conversation session identifier.
-            lang: BCP-47 language code.
-            units: Preferred measurement system (``"metric"`` / ``"imperial"``).
-
-        Returns:
-            The assistant's final response after the loop has completed.
-        """
-        raise NotImplementedError()
-
-
 class SummarizerEngine(AbstractAgentEngine):
     """Engine designed for condensing long documents into concise summaries."""
 

ovos_plugin_manager/utils/__init__.py

@@ -99,7 +99,6 @@ class PluginTypes(str, Enum):
     WEB_PLAYER = "opm.media.web"
     PERSONA = "opm.plugin.persona"  # personas are a dict, they have no config because they ARE a config
     AGENT_TOOLBOX = "opm.agents.toolbox"
-    AGENT_LOOP = "opm.agents.loop"
 
     # solver plugins are deprecated!
     QUESTION_SOLVER = "opm.solver.question"
@@ -154,7 +153,6 @@ class PluginConfigTypes(str, Enum):
     AGENT_COREF = "opm.agents.coref.config"
     AGENT_YES_NO = "opm.agents.yesno.config"
     AGENT_TOOLBOX = "opm.agents.toolbox.config"
-    AGENT_LOOP = "opm.agents.loop.config"
     KEYWORD_EXTRACTION = "opm.keywords.config"
     UTTERANCE_SEGMENTATION = "opm.segmentation.config"
     TOKENIZATION = "opm.tokenization.config"